© CDISC 2013
Presented by
Kunihito Ebi, Senior Consultant at Fujitsu
2
Feasibility Study on Applying CDISC SDTMto Post-Marketing Surveillance
© CDISC 2013
Contents
• Applying SDTM to Post-Marketing Surveillance
What it means
Values of SDTM in PMS
• Technical Approach and Considerations
Reliability and Trustworthiness of Records
Cost Effectiveness
3
© CDISC 2013
What is Post-Marketing Study
• Post-Marketing Studies in Japan are regulated byGood Post-marketing Study Practice (GPSP) Phase VI Clinical Study
“使用成績調査” – Study of real-life uses of a new drug-> This is the topic of my presentation.
• Study of real-life uses of a new drug Approximately 3,000 subjects
No placebo/comparators
Typically a specialized EDC and/or a paper CRF areused to capture patient records
5
© CDISC 2013
Why Applying SDTMto Post-Marketing Studies?
• Accelerate Studies and Reduce Costs As fast and economy as clinical trials
Promote use of EDC for PMS
Standardize analytical datasets and programs
• Better comparison between clinical trial and PMS Enhance data comparability between well-controlled
clinical studies and real-life PMS
Enable sharing methods/resources in clinical and PMS
• Enabler of Healthcare-Link Accelerate speed and improve data quality
Enable collecting more valuable information
6
© CDISC 2013
Technical Architecture
• Tools EDC – PostMaNet
ETL – Clinical Service Bus
8
PostMaNet
Export
Clinical Service Bus
MappingSpecification
Document
SDTMDatasets
Define.xml
© CDISC 2013
Key Challenges
• How can we ensure that SDTM datasets are asreliable/trustworthy as PostMaNet source datawhile minimizing manual QCs on the SDTMdatasets?
• To what extent can we standardize datatransformation programs in order to minimize costfor SDTM transformation per protocol?
• These are universal challenges with not limitedto PMS data but including clinical trials data.
9
© CDISC 2013
Reliability and Trustworthiness –Our Understanding
• SDTM datasets that are relied on to performregulated activities or those submitted to aregulatory agency are considered to be subject toElectronic Records requirements in the region.
• A sponsor should make a decision on the extentof the validation taking into account the impacton the reliability and trustworthiness of therecords.
10
© CDISC 2013
Reliability and Trustworthiness –Validation Approach
• Mapping Specification Document Including mapping between source system to SDTM,
coding conversions, derivation formula, and so forth.
• Program Design & Test Needed for each transformation program (could be for
each transformation rule)
• Execution Records Logs may substitute if SDTM datasets are produced in a
fully systematic manner
• Qualification of SDTM Datasets OpenCDISC may validate conformance to SDTM, but
does not qualify all mapping specifications are met.
Manual QC may still be needed.
11
© CDISC 2013
Transformation Steps
12
PostMaNet Data
Step 1: Split into CSV files per PostMaNet variable
Step 2: Convert value of each variable
Step 3: Map variables to SDTM variables
Split Files
Converted Files
SDTM Datasets
MappingSpecifications
MappingSpecifications
Log
Log
Log
© CDISC 2013
Reliability and Trustworthiness –Audit Trails
• A typical audit trail record contains who, when,and what (e.g. old value/new value) information.
• What are audit trails in terms of the followingSDTM transformation? when {0, 1} values are converted to {M, F}
when the minimum Exposure Date on CRF is calculatedto derive SDTM Subject Reference Start Date
when a field on CRF is mapped to a different SDTMvariable based on its context
13
© CDISC 2013
Reliability and Trustworthiness –Our Logging Approach
• We identified that the following information shouldappear in the transformation log; Execution user name
Date/time of execution
Input/output file names (i.e. variable names)
Number of records processed
• The logs will need to be retained along withinput/output data during the required recordretention period.
14
© CDISC 2013
Cost Effectiveness – Our Findings
• Many transformation programs can be reused Programs at step1 & 3
Common patterns of transformations at step 2
• SDTM transformation tend to require ad-hocprograms thus increase the cost, when; Variable names and/or codelists in the source system
are inconsistent across protocols.
Mapping is determined by the value of data
SUPP-- domains are used.
* Above is only a partial list.
15
© CDISC 2013
Summary
• SDTM can be applied to post-marketing studiessimilarly to clinical studies.
• Maintaining reliability/trustworthiness andincreasing cost effectiveness are the two keychallenges in implementing transformation.
• Validation and logging approach should bedefined clearly to ensure regulatory compliance.
• Use of ad-hoc programs should be minimized toreduce costs.
16
© CDISC 2013
Thank you for listening!
Please feel free to send your questions to:
Kunihito Ebi, Senior Consultant
Fujitsu Limited.
17